Should Ethereum split forks into smaller scopes and ship them more often?
Governance & community · 10 participants · 13 quotes · 2026-11-20 ~ 2026-11-20
When the Glamsterdam delay pushed back features like contract size expansion, app developers such as lex_node and sakulstra called for shipping smaller forks more frequently or shipping a lead fork containing only quality-of-life (QOL) features first. Core developers including lightclients and CPerezz19 agreed smaller forks ship faster but pushed back on the practicality, citing the cost of scope cuts, retesting, and the bottleneck of protocol understanding.
Background
Glamsterdam is the largest hard fork to date, bundling over 20 EIPs including ePBS, BALs, and EIP-8037. It originally targeted mainnet in early 2026, but a problem surfaced where EIP-8037 breaks invariants of already-deployed contracts, pushing the earliest mainnet date to mid-January 2027. According to nero_eth's analysis, every fork since Dencun has taken more than 400 days from scope finalization to mainnet, but Glamsterdam targeted the shortest timeline yet at 320 days. Pectra split its fork and Fusaka dropped EOF at the last minute, but Glamsterdam reached testnet while keeping nearly all its EIPs, and even after the public testnet there were 10 further spec changes over six weeks.
Why it matters
The tension between app developers' demand for faster shipping and the limits of core developers' review and comprehension capacity could change how scope is decided for the next fork and the practice of agreeing on cut dates in advance. This connects to the roadmap prioritization debate, since prolonged contract size constraints increase app developers' incentive to migrate to their own L2s.
Key points
- Can a small lead fork (Proto-glamsterdam) containing only QOL features like contract size expansion be shipped first?
- Can AI raise fork throughput enough to allow more frequent forking, or is the work of understanding the protocol itself the bottleneck?
- Is cutting scope itself, as with dropping FOCIL, even a viable decision?
For — Split forks into smaller scopes and ship more often
sakulstra, _gabrielShapir0, cyp.eth, Łukasz Rozmej, Philastru
- Asks whether coordination is really that hard, given that smaller-scope hard forks could simply be shipped more often (sakulstra).
- Since the most painful part of the Glamsterdam delay is the contract size expansion delay, ship a small Proto-glamsterdam containing only QOL features first (lex_node).
- If hard forks ship on a 6 to 7 month cycle, validators can keep up (0xcyp).
- Work has shifted from writing to review and guidance, which takes time, but it can be parallelized far more, raising throughput (URozmej).
Against — It's impossible, or actually more work
ً, 0xRahul, CPerezz.eth ⟠ 🦀
- The chance of shipping Proto-glamsterdam first is 0% (lightclients).
- Making a separate new fork actually means more work, since it requires removing code and retesting (omw_to_the_moon).
- ACD even did research backing this, but I still don't see why this is such a big problem; expanding beyond 64kb is much harder (CPerezz19).
Neutral — Smaller forks do ship faster, but scope cuts and comprehension are the bottleneck
Toni Wahrstätter ⟠
- As past cases like Fusaka show, smaller forks really do ship faster.
- Work to understand the protocol can be accelerated by AI, but it cannot be parallelized.
- Decisions to cut scope, like whether to drop FOCIL to ship Frame faster, are themselves hard to make.
Other — Delay doesn't matter, only uptime matters
hammundi7
- If you believe in Ethereum, you should believe that only full uptime matters, not delays and schedules.